Devices that surpass humans with 73% accuracy and expose their lies

Devices that surpass humans with 73% accuracy and expose their lies

For decades scientists have been trying to come up with an accurate way to find out if someone is lying or cheating on a target, and now, thanks to artificial intelligence, scientists think they are a significant percentage closer to achieving the goal.

Artificial intelligence can detect deception

According to the British Daily Mail, citing the leading journal Brain and Behavior, a team of scientists has succeeded in creating sensors that can read the faces of many volunteers and monitor and observe small changes in facial movements. When they tell false and true facts.

73% accuracy

The artificial intelligence system was able to tell if the lie was 73% accurate, which is slightly lower than the polygraph test, which reached 80% accuracy, but scientists say this percentage was achieved much earlier. Phase of the experiment, i.e. it will improve in the future.

Scientists at Tel Aviv University hope to use AI-equipped cameras in the future to see if anyone is lying at airports, online job interviews or police interrogations of suspects.

Tools are better than humans

Previous research has shown that humans can lie up to 55% of the truth, while a polygraph is up to 80% accurate, but this percentage is not enough to accept the results of a lie detection test. Court, so researchers around the world are working for new solutions.

The researchers used machine learning and artificial intelligence programs to quickly analyze small changes in muscle movements while lying down, including small movements of the cheek muscles and eyebrows.
The amount of torsion of the facial muscles.

See also  Internet: Internet and mass media: Lenta.ru

The facial movements were measured using stickers printed on soft surfaces containing electrodes capable of monitoring and measuring nerves and muscles.

The study is based on the assumption that facial muscles twist when a person lies, and that no electrodes are yet sensitive to measure this torsion, so the researchers trained a machine learning program to detect lies based on EMG signals from the electrodes. By applying this method, accuracy of up to 73% is achieved, which may not be ideal, but it is far superior to other technologies used for the same purpose.

Dramatic effects

The researchers hope that the advanced AI program will have “dramatic results” as more electrodes become available in the future.

Thus “advanced technology can be used in areas such as interrogation in banks, police departments, airports or online job interviews. [إذ] High resolution cameras trained to detect facial muscle movements [تحديد ما إذا كانت] Honest statements from lies. “

Criticisms of the past

The idea of ​​using artificial intelligence to detect fakes has been criticized before. For example, in 2018, researchers at Manchester Metropolitan University suggested that AI could be used to find out if people are lying to border security by analyzing their precise gestures, but Ray Paul, a professor of criminal investigation at the University of Derby, criticized the idea. , Said it was a project that was not credible because observing small changes in people’s faces was not proof that it was an accurate way to measure lies.

Professor Ball said: “They [علماء جامعة مانشستر متروبوليتان] They deceive themselves that they will ever be very effective. They waste a lot of money because that technology is based on the basic misconception that human behavior can be honest and deceptive.

See also  Delete Account: How to Avoid WhatsApp Privacy Policy If You Do Not Like It, Learn the Details - Delete Your Account Step by Step If You Are Not Satisfied With The New Privacy Policy

Written By
More from Josh Atkinson
Sony Reiterates Its Aim to Increase Very first-Party Titles to Personal computer to Increase Profitability
The release of Horizon Zero Dawn Comprehensive Edition for Laptop, one of...
Read More
Leave a comment

Your email address will not be published. Required fields are marked *